Neil Rothwell – Director The opportunities for the use of R in real time industrial analytics Dr Neil Rothwell C3global Limited
Neil Rothwell – Director What is this all about Operational/Industrial analytics Why now Is there a market? Barriers to adoption Is it really that simple? Where does R fit
Neil Rothwell – Director IT and OT have evolved from totally different start points and have been based on different data types and user needs IT has developed from the top down and focussed on business and corporate requirements. Thus the needs of the engineer are rarely met. OT has developed from the bottom up with control of proprietary equipment being the prime goal. These systems have low tolerance to failure so are not deigned to be flexible. Hence the barrier to integration that has led to low takeup Why Now? - ‘IT OT Convergence’
Neil Rothwell – Director ERP, CRM, Financial, Scheduling, Resource planning have all been delivered If we are to deliver further savings in industrial applications need to address the operation In the North Sea effectiveness/efficiency is ~60% with ambition to reach 80% Common need across all industries CIOs are changing Why Now? – Done everything else
Neil Rothwell – Director Neil Rothwell – Technical Director 5 Growth of Industrial Operational Technologies Source: Defining and Sizing the Industrial Internet, Wikibon 2013 Projected Spend & Value Delivered by the Industrial Internet (USD Billions) Value Delivered by Industrial Internet Technology Investments Total Industrial Internet Technology Spend $1,400 $1,200 $1,000 $800 $600 $400 $ $514 Bn $1,279 Bn $20 Bn $23 Bn Industrial Internet spend will grow at a 50% CAGR to reach $514Bn Analytics focus for Fortune 500 moving to industrial operations: “Analytics have been transformative in wide areas of customer & product service…industrial analytic applications are the next frontier.” Operational technologies include: Core systems that deal with running plant and industrial equipment. Devices and sensors to ensure system integrity and to meet technical constraints. Event-driven and frequently real- time software applications or devices with embedded software. Systems used to manage and control mission-critical production or delivery processes.
Neil Rothwell – Director 6 Need to bridge the gap Between IT and OT Operational Technologies: “Exist in all industries & are a major source of big data with massive data volumes, disparate sources & variable latency.” “There are few companies capable of providing applications with embedded predictive analytics” to predict asset performance and optimize operational production. Costs Orders Constraints Specifications State Location Conditions Events Context Visualization Prediction Monitoring ERP CRM Demand Management Financial Planning Network Optimization Control Systems Smart Machines Location Systems Environment Sensors Barcodes / RFID Operational TechnologyInformation Technology
Neil Rothwell – Director Need Both Predictive & Prescriptive Analytics Need to Monitor Assets & Events in Real-time to Anticipate Operational Needs Anticipate events Identify event correlations and leading indicators Create models for propensity and risk Forecast performance Descriptive : What happened? Diagnostic : Why did it happen? Predictive : What will happen? Prescriptive : What should I do? Decision Support Decision Automation Data DecisionAction 7 A need to anticipate future events that are often missed by traditional BI tools and gives operators forward looking action plans to mitigate operational risk Process Flow -> From Raw Data to Actionable Operational Intelligence
Neil Rothwell – Director Multiple data sources Time series, metadata, batch Structured data needs to be used in an unstructured manner Traditional systems, SCADA, DCS, historians not ideal toolsets for data manipulation Solutions tend to be domain and vertical specific so niche Barriers to adoption
Neil Rothwell – Director Neil Rothwell – Technical Director Industry doesn’t even figure in the research on maturity
Neil Rothwell – Director Industrial Analytics Maturity
Neil Rothwell – Director Privileged and Confidential – © C3global Ltd., All Rights Reserved 11 Case Study– from regression to total asset management South Australia is the driest State in the driest continent on earth The main source of water is the Murray River which is dependent on inflows from the North and East of Australia The last period of drought prompted the building of a high capacity desalination plant and a North/South interconnecting pipeline which, for the first time, allowed SA Water to move water almost anywhere in their network This new infrastructure and the ongoing investment in telemetry equipment lead to SA Water looking for smarter ways to manage their network and optimize water usage
Neil Rothwell – Director Privileged and Confidential – © C3global Ltd., All Rights Reserved 12 Demand Forecasting Water demand varies due to a number of factors including: Population Residential development Industrial activity Temperature Rainfall Water stock levels Water usage restrictions Weather patterns and weather events Tourism, major entertainment and sporting events (in the hot months of February and March the population of Adelaide can increase by hundreds of thousands) Maintenance activity Large scale infrastructure projects
Neil Rothwell – Director Privileged and Confidential – © C3global Ltd., All Rights Reserved 13 The Solution System uses data from both the IT and OT systems to perform complex calculations providing demand forecasts on an hourly, daily, weekly, monthly and annual basis. Long term prediction of demand over a 10 year time frame The major systems to which we interface include: Production Data Customer Billing Bureau of Meteorology Water Quality Maintenance Activity Water Availability Geospatial Information Demand Areas and Usage Patterns
Neil Rothwell – Director Privileged and Confidential – © C3global Ltd., All Rights Reserved 14 Observed versus Predicted
Neil Rothwell – Director Privileged and Confidential – © C3global Ltd., All Rights Reserved 15 Forecast and predicted values compared in model
Neil Rothwell – Director Privileged and Confidential – © C3global Ltd., All Rights Reserved 16 Demand forecasting by sector, customer, zone, season etc.
Neil Rothwell – Director Privileged and Confidential – © C3global Ltd., All Rights Reserved 17 Summary and detailed analysis including scenarios
Neil Rothwell – Director Privileged and Confidential – © C3global Ltd., All Rights Reserved 18 The Outcome Weather forecast leads to demand forecast From demand forecast can do production planning From production planning can determine where to produce the water Transportation costs then come into play Energy tariff then determine when you make it This then determines the energy requirements Energy requirement can be determined in advance Energy can be bought on spot market December 2013 temperatures reached 47C, produced minimal requirement in water and saved over $1.3M in a week
Neil Rothwell – Director Water Quality
Neil Rothwell – Director Techniques used Multivariate k-Nearest Neighbour Clustering Algorithm Binomial Event Detection Consecutive Sampling Probability Binary Tree
Neil Rothwell – Director Event Detection - From raw data to event probabilities Gather Data Alarm Aggregate Predict Filter
Neil Rothwell – Director Station Overview
Neil Rothwell – Director Privileged and Confidential – © C3global Ltd., All Rights Reserved 23 Solution Examples by Industry Predict corrosion rates in pipelines, downholes, etc. Predict scale and well intervention requirements Predict chemical consumption Predict production flow rates and usage per year Integrity operating window & dashboard Predict corrosion rates in pipelines, downholes, etc. Predict scale and well intervention requirements Predict chemical consumption Predict production flow rates and usage per year Integrity operating window & dashboard Predict cable and hot joint condition Anticipate transformer and compressor failure through real-time health monitoring Provide tower refurbishment & maintenance plan based on corrosion rate modeling Predict tree maintenance near overhead lines Predict cable and hot joint condition Anticipate transformer and compressor failure through real-time health monitoring Provide tower refurbishment & maintenance plan based on corrosion rate modeling Predict tree maintenance near overhead lines Predict water production levels and consumer demand Forecast effects on demand based on temperature patterns Alarm suppression based on predicted weather patterns across network Prescribe optimal water movements based on cheaper tariffs Predict water production levels and consumer demand Forecast effects on demand based on temperature patterns Alarm suppression based on predicted weather patterns across network Prescribe optimal water movements based on cheaper tariffs Energy consumption modeling and targeting Predict and manage carbon footprint & output Predict fuel usage to optimize fleet management Energy consumption modeling and targeting Predict and manage carbon footprint & output Predict fuel usage to optimize fleet management Anticipate failure of refrigeration equipment Prescriptive load shedding for optimal power reduction Prescriptive set point remapping when overriding settings Identify and prescribe field service maintenance needs Anticipate failure of refrigeration equipment Prescriptive load shedding for optimal power reduction Prescriptive set point remapping when overriding settings Identify and prescribe field service maintenance needs Oil and GasPower TransmissionWater & UtilitiesSustainabilityRefrigeration
Neil Rothwell – Director We have just started to understand the power of analytics in operational environments There will be a rapid growth in the use of R in real time industrial environments